With recent updates to AI models like ChatGPT, several users have reported experiencing a sense of “laziness” from the chatbot when performing various tasks. This issue, which appeared to arise from the GPT-4 model, was eventually acknowledged by OpenAI, the creator of ChatGPT.
OpenAI stated on December 11, 2023, “We’ve heard all your feedback about GPT-4 becoming lazier! We haven’t updated the model since November 11, and it’s certainly not intentional. The model’s behavior can be unpredictable, and we’re looking to fix it.“
Introduced in March 2023, GPT-4 is a language model used by ChatGPT, which has become widespread, including in Microsoft’s products and services—a key partner of OpenAI. This was followed by another significant update, GPT-4 Turbo, in November 2023. Besides enhanced capabilities, this version accesses a much more recent database, allowing it to respond better to current topics.
However, both of these updates have been criticized for ChatGPT’s perceived lack of effort in certain situations, with users questioning if GPT-4 had become less intelligent or even regressed compared to GPT-3.5.
On forums like Reddit, users have shared experiences of ChatGPT’s unexpected responses, such as suggesting users complete tasks themselves, which seemed basic for the AI. These incidents have sparked discussions and theories about the reasons behind this lack of enthusiasm and strategies to work around the problem.
OpenAI, in response to these concerns, included a fix in their January 25, 2024, update notes. The new version, called gpt-4-0125-preview and specific to GPT-4 Turbo, aims to reduce instances of ‘laziness’ where the model fails to complete a task. This patch is available for users with an active subscription to GPT-4 Turbo.
While it may take time to observe the differences between gpt-4-0125-preview and older versions, even if the patch is effective, there’s no guarantee that similar issues won’t reappear in future updates of GPT-4 or in new language models.
